Eclipse: Disambiguating Illumination and Materials using Unintended Shadows (2305.16321v3)
Abstract: Decomposing an object's appearance into representations of its materials and the surrounding illumination is difficult, even when the object's 3D shape is known beforehand. This problem is especially challenging for diffuse objects: it is ill-conditioned because diffuse materials severely blur incoming light, and it is ill-posed because diffuse materials under high-frequency lighting can be indistinguishable from shiny materials under low-frequency lighting. We show that it is possible to recover precise materials and illumination -- even from diffuse objects -- by exploiting unintended shadows, like the ones cast onto an object by the photographer who moves around it. These shadows are a nuisance in most previous inverse rendering pipelines, but here we exploit them as signals that improve conditioning and help resolve material-lighting ambiguities. We present a method based on differentiable Monte Carlo ray tracing that uses images of an object to jointly recover its spatially-varying materials, the surrounding illumination environment, and the shapes of the unseen light occluders who inadvertently cast shadows upon it.
- Differentiable rendering of neural SDFs through reparameterization. SIGGRAPH Asia, 2022.
- Inferring light fields from shadows. CVPR, 2018.
- Shape, illumination, and reflectance from shading. TPAMI, 2015.
- Neural reflectance fields for appearance acquisition. arXiv, 2020a.
- Deep Reflectance Volumes: Relightable Reconstructions from Multi-View Photometric Images. ECCV, 2020b.
- NeRD: Neural reflectance decomposition from image collections. ICCV, 2021.
- Turning corners into cameras: Principles and methods. ICCV, 2017.
- JAX: composable transformations of Python+NumPy programs, 2018. http://github.com/google/jax.
- Paul Debevec. Rendering synthetic objects into real scenes: Bridging traditional and image-based graphics with global illumination and high dynamic range photography. SIGGRAPH, 1998.
- Reconstructing Translucent Objects using Differentiable Rendering. SIGGRAPH, 2022.
- Computing Fourier Transforms and Convolutions on the 2-Sphere. Advances in applied mathematics, 1994.
- Learning to predict indoor illumination from a single image. ACM Trans. Graph., 2017.
- Deep outdoor illumination estimation. CVPR, 2017.
- Dr.Jit: A Just-In-Time Compiler for Differentiable Rendering. SIGGRAPH, 2022.
- Real shading in unreal engine 4. Proc. Physically Based Shading Theory Practice, 2013.
- Adam: A Method for Stochastic Optimization. ICLR, 2015.
- Neroic: Neural rendering of objects from online image collections. ACM Trans. Graph., 2022.
- Stanford-ORB: A real-world 3d object inverse rendering benchmark. NeurIPS, 2023.
- Estimating the natural illumination conditions from a single outdoor image. IJCV, 2012.
- Deeplight: Learning illumination for unconstrained mobile mixed reality. CVPR, 2019.
- Learning illumination from diverse portraits. SIGGRAPH Asia 2020 Technical Communications, 2020.
- Differentiable monte carlo ray tracing through edge sampling. SIGGRAPH Asia, 2018a.
- Learning to reconstruct shape and spatially-varying reflectance from a single image. SIGGRAPH Asia, 2018b.
- Inverse rendering for complex indoor scenes: Shape, spatially-varying lighting and svbrdf from a single image. CVPR, 2020.
- NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis. ECCV, 2020.
- Instant neural graphics primitives with a multiresolution hash encoding. SIGGRAPH, 2022.
- Extracting Triangular 3D Models, Materials, and Lighting From Images. CVPR, 2022.
- Mitsuba 2: A retargetable forward and inverse renderer. SIGGRAPH Asia, 2019.
- Radiative Backpropagation: An Adjoint Method for Lightning-Fast Differentiable Rendering. SIGGRAPH, 2020.
- Eyes for relighting. ACM Trans. Graph., 2004.
- Seeing the world in a bag of chips. CVPR, 2020.
- Physically Based Rendering: From Theory to Implementation (3rd ed.). Morgan Kaufmann Publishers Inc., San Francisco, CA, USA, 3rd edition, 2016.
- A signal-processing framework for inverse rendering. SIGGRAPH, 2001.
- Illumination from shadows. TPAMI, 2003.
- Neural illumination: Lighting prediction for indoor environments. CVPR, 2019.
- NeRV: Neural reflectance and visibility fields for relighting and view synthesis. CVPR, 2021.
- Objects as cameras: Estimating high-frequency illumination from shadows. ICCV, 2021.
- Fourier features let networks learn high frequency functions in low dimensional domains. NeurIPS, 2020.
- Accidental pinhole and pinspeck cameras: Revealing the scene outside the picture. CVPR, 2012.
- Optimally Combining Sampling Techniques for Monte Carlo Rendering. SIGGRAPH, 1995.
- Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields. CVPR, 2022.
- Differentiable Signed Distance Function Rendering. SIGGRAPH, 2022.
- Microfacet models for refraction through rough surfaces. Eurographics, 2007.
- Using unknown occluders to recover hidden scenes. CVPR, 2019.
- Monte Carlo Estimators for Differential Light Transport. SIGGRAPH, 2021.
- PhySG: Inverse rendering with spherical gaussians for physics-based material editing and relighting. CVPR, 2021a.
- NeRFactor: Neural factorization of shape and reflectance under an unknown illumination. SIGGRAPH Asia, 2021b.
- Dor Verbin (21 papers)
- Ben Mildenhall (41 papers)
- Peter Hedman (21 papers)
- Jonathan T. Barron (89 papers)
- Todd Zickler (23 papers)
- Pratul P. Srinivasan (38 papers)